Replication Archive for

Coppock, Alexander. Avoiding Post-Treatment Bias in Audit Experiments
Journal of Experimental Political Science, Forthcoming

This replication archive contains three files:

1. voterIDexp_data_sept2014.RData, which is the .rdata available from http://dx.doi.org/10.7910/DVN/28158, the replication archive for White et al. (2015)

2. aec_replication_JEPS.R, the script that produces table 1

3. aec_replication_JEPS.html, which is a log file from a run of aec_replication_JEPS.R

The variables used in the analysis are:

first_name_latino: binary treatment indicator
bounced: binary indicator for whether the email bounced. I condition on emails not bouncing, which is ok because bouncing is independent of treamtent assignment
response: the raw binary variable for whether the email received a response (includes NAs)
response_no_na: the cleaned binary variable that codes "NA" as not receiving a response
friendly: the binary variable that indicates if the original researchers coded the response as friendly or not (includes NAs)
friendly_no_na: this variable imputes a 0 if "friendly" is NA
pair_id: an indivator for the matched pair. (The original researchers did the matching)
AR: a logical indicating that the unit belongs to a pair in which both members responded

All cleaning and manipulation of the replication dataset for White et al. is conducted in aec_replication_JEPS.R


